bowtie vs bowtie2 for RSEM analysis trinity assembled transcripts

[horvathdp]

I have a Trinity assembly from a time course experiment in a non-model plant. The assembly seems excellent, and I am getting ready to use RSEM to look for deferentially expressed genes. Bowtie2 gives me a lot more alignments ~75% of my reads aligning) than Bowtie (about 26%), but of those that align with Bowtie, nearly 100% (about 3.5 million reads) correctly map with their paired ends (relative to only about 16% or 2.1 million for the Bowtie2 mapping).

I am worried that the Bowtie2 alignment may be giving me too many false positive hits because they are mapping to assembled transcripts that did not come from the given sample.

Is bowtie or bowtie2 recommended for alignment for RSEM expression analysis?

Dave

[westerman]

I believe that bowtie is recommended. At least the Trinity pipeline still uses it. If you want to use Bowtie2 then using the parameters:

--no-discordant --no-mixed -X 1200

Should keep bowtie2 from mapping reads to different Trinity transcripts.